SeC: Advancing Complex Video Object Segmentation via Progressive Concept Construction

[📂 GitHub] [📦 Model] [🌐 Homepage] [📄 Paper]

Highlights

• 🔥We introduce Segment Concept (SeC), a concept-driven segmentation framework for video object segmentation that integrates Large Vision-Language Models (LVLMs) for robust, object-centric representations.
• 🔥SeC dynamically balances semantic reasoning with feature matching, adaptively adjusting computational efforts based on scene complexity for optimal segmentation performance.
• 🔥We propose the Semantic Complex Scenarios Video Object Segmentation (SeCVOS) benchmark, designed to evaluate segmentation in challenging scenarios.

SeCVOS Benchmark

We propose the Semantic Complex Scenarios Video Object Segmentation (SeCVOS) benchmark, specifically designed to assess a model’s ability to perform high-level semantic reasoning across complex visual narratives. SeCVOS contains 160 carefully curated multi-shot videos characterized by: 1) Highly discontinuous frame sequences, 2) Frequent reappearance of objects across disparate scenes, and 3) Abrupt shot transitions and dynamic camera motion.

License

Our annotations are licensed under a CC BY-NC-SA 4.0 License. They are available strictly for non-commercial research.

We uphold the rights of individuals and copyright holders. If you are featured in any of our video annotations or hold copyright to a video and wish to have its annotation removed from our dataset, please reach out to us. Send an email to zhangzhixiong@pjlab.org.cn with the subject line beginning with SeCVOS, or raise an issue with the same title format. We commit to reviewing your request promptly and taking suitable action.

---

Citation

If you find this project useful in your research, please consider citing:

@article{zhang2025sec,
  title     = {SeC: Advancing Complex Video Object Segmentation via Progressive Concept Construction},
  author    = {Zhixiong Zhang and Shuangrui Ding and Xiaoyi Dong and Songxin He and Jianfan Lin and Junsong Tang and Yuhang Zang and Yuhang Cao and Dahua Lin and Jiaqi Wang},
  journal   = {arXiv preprint arXiv:2507.15852},
  year      = {2025}
}